Due to the similar order of magnitude of the defect size and the detection wavelength, when detecting micro-/nano-scale defects on the surface of a fine optical component, the intense modulation of the optical field poses challenges in correlating imaging widths of defects with actual widths. A dark-field scattering imaging (DFSI) model, based on the finite difference time domain (FDTD) method, is established to study the imaging for triangular and circular section defects and investigate the influence of wavelength on defect width detection. Simulated results indicate that a shorter wavelength of the light source in a DFSI detection system leads to a larger mapping range between the imaging width and the actual defect width, which makes calibration less difficult. A DFSI detection system for micro-/nano-scale surface defects on optical components is built to test defects with rectangular cross-sections on a calibration board. The defect widths estimated from the experimental and simulated results are in good agreement, with a root-mean-square error (RMSE) of 0.11 µm.